Word2Vec.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Word2Vec Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import gensim\n",
    "\n",
    "from utils import *\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings(action='ignore', category=UserWarning, module='gensim')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.1. Train Word2Vec Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Getting Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# I use a memory friendly iterator\n",
    "text_w2v = MySentences(dirname='./Data/')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### CBOW Version"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model_w2v = gensim.models.Word2Vec(sentences=text_w2v, size=300, sg=0, hs=1)\n",
    "if not os.path.exists('./WE_models'):\n",
    "    os.mkdir('./WE_models')\n",
    "model_w2v.save('WE_models/w2v_cbow_300D')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Skip-Gram Version"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_w2v = gensim.models.Word2Vec(sentences=text_w2v, size=300, sg=1, hs=1)\n",
    "if not os.path.exists('./WE_models'):\n",
    "    os.mkdir('./WE_models')\n",
    "model_w2v.save('WE_models/w2v_sg_300D')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1.2. Load Word2Vec Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### - CBOW Version"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_w2v = gensim.models.Word2Vec.load('WE_models/w2v_cbow_300D')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- #### Similar Words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('facile', 0.7146144509315491),\n",
       " ('dur', 0.675102710723877),\n",
       " ('pénible', 0.6174148321151733),\n",
       " ('agréable', 0.6010310649871826),\n",
       " ('amusant', 0.5417053699493408),\n",
       " ('étrange', 0.5337499976158142),\n",
       " ('impossible', 0.5333918929100037),\n",
       " ('triste', 0.5296621322631836),\n",
       " ('gênant', 0.5262279510498047),\n",
       " ('important', 0.5219485759735107)]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_w2v.wv.most_similar('difficile')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### - Skip-Gram Version"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model_w2v = gensim.models.Word2Vec.load('WE_models/w2v_sg_300D')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- #### Similar Words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('dur', 0.5741609334945679),\n",
       " ('pénible', 0.5671994686126709),\n",
       " ('rude', 0.5307800769805908),\n",
       " ('facile', 0.5286517143249512),\n",
       " ('éprouvante', 0.5169097185134888),\n",
       " ('éprouvant', 0.5106732845306396),\n",
       " (\"s'imaginer\", 0.5098478198051453),\n",
       " ('instructif', 0.49842825531959534),\n",
       " ('stressante', 0.49489402770996094),\n",
       " ('frustrant', 0.4873944818973541)]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_w2v.wv.most_similar('difficile')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}